I do not really know, but my general guess would be that at high frequencies (but still below optical transitions band) you might be doing something to the electron band structure of the ZnO ? if the oscillator strength of the optical transitions goes down, so will the resulting "dielectric constant".
To proceed further, you have to tell me what is the band gap of ZnO and how stable it is as you increase the temperature?
Also, is it really dielectric constant you measure or is it just the tails of the dielectric response function (weak frequency dependences close to the various absorption bands)?
You have right sir, it's almost constant. I make a verifcation and we can consider that the dielectric constant at high energy is independent to temperature.